XPeng’s Focus on Autonomous Driving and Robotaxi in 2022

XPeng’s peak in the 1024 news cycle has passed. Due to the industry’s internal competition, there are often reports about the company that disregard crucial details due to the pressure of time. Thus, it is crucial to point some details out.

One page of the company’s presentation stated that “XPeng aims to become the first Chinese auto company to explore autonomous driving mobility service operation by H2 2022.”

Although CEO He XPeng did not say much about it during the press conference, he elaborated on this topic during an interview after.

It is highly probable that the company is referring to Robotaxi services (which He confirmed later in the interview), but there are two probable routes to take when developing this product.

The more common route is to establish a single-purpose autonomous driving company, similar to Waymo, Baidu, and Xiaoma Zhixing, with the aim of commercializing Robotaxi services. Nonetheless, there are not many hopeful signs pointing towards these companies’ profit-making from this industry yet.

Many people in China also hold a pessimistic view on this development path, as demonstrated by one widely circulated opinion: “Companies that primarily focus on Robotaxi services will ultimately fail.”

Thus, Tesla is currently the only traditional carmaker that has stated its intention to explore autonomous driving mobility service operations. But before this is possible, a company must first attain L5 autonomous driving capabilities, which Tesla has yet to achieve.

In light of the unclear nature of this business model, what is XPeng’s ultimate goal in exploring autonomous driving mobility service operations?

He XPeng provided an answer during the interview, saying: “It is similar to Robotaxi services’ core logic, in which the core idea is to have a service that operates within a specific area of a city, but where safety officers are present who accumulate data, particularly data on any exceptions.

For example, data on road closures, which is unique data or data that can arise from unexpected weather events; these will never be seen under normal circumstances. It is only through this data accumulation, which can potentially take millions of accumulated miles and a significant amount of time, that larger-scale projects such as autonomous driving or even unmanned driving can become a reality.

Ultimately, the most important issue is the robustness of the system. The service may pretend to tread the same path, have the same production processes, the same cost-efficiency, and satisfactory safety levels, but for it to be truly robust and operate well even in chaotic, unexpected scenarios is one hundred times harder. Zhong Hua, our recent partner, emphasizes two essential loops, and these loops must be interconnected and must also run an extensive amount of data for the service to function.”

Thus, it is evident that XPeng plans to start laying the groundwork for Robotaxi services in 2022, but its primary goal is to gather data and to improve the robustness of its autonomous driving system, which is a common goal shared by all companies working on this technology.In the XPILOT section of the Science and Technology Day, Dr. Xinzhou Wu shared several relevant pieces of information.

Firstly, the positioning method of Xpeng’s NGP has shifted from mainly GPS to mainly vision and IMU. Currently, Xpeng relies heavily on GPS positioning for high-speed NGP, with the high-precision GPS positioning of the high-precision map being similar to that on smartphones, only with higher accuracy. With GPS positioning, Xpeng achieved navigation-assisted driving on the highway.

However, city roads differ from highways, with many tunnels, buildings, and viaducts, making it easy to lose GPS positioning. If GPS positioning is lost, navigation is naturally inaccurate, as many people have experienced in smartphone navigation. In order to solve the problem of city positioning, Xpeng has chosen to use vision + IMU positioning that is independent of signals. The advantage is that it is not bound by GPS signals, but the downside is that the difficulty of visual positioning is extremely high, and the current visual perception of distance measurement is generally poor, so it is a big challenge to achieve accurate distance measurement with vision alone.

In addition, the goal is to update high-precision map data on a daily basis.

Secondly, Dr. Wu believes that data loop + map loop are essential for intelligent driving in the latter half of the game. It is expected that by the end of 2022, the mileage of high-speed NGP will exceed 120 million kilometers and that of city NGP will be 35 million kilometers.

Finally, here are some Q&A regarding the interview:

About Xpeng brand positioning

Q: When it comes to key brand positioning words and brand sharpening, if we don’t give very clear keywords to outline Xpeng’s image, what kind of image will consumers have when they mention Xpeng?

A: We will hold a small brand launch event next month, and then we will share it with everyone.

Q: I am curious about your five-year or ten-year vision. What kind of company will Xpeng be in five or ten years? To what extent do you think Xpeng has met your expectations so far?

A: We will hold a launch event next month to talk about our brand vision, and I will give some answers at that time. The other part of the answer is about the ten-year vision; to be honest, I have a macroscopic idea about it and don’t have a specific framework. The reason why I cannot think of it is that technological change is too fast and beyond my imagination.

About supercharging deployment considerations

Q: We released the 800V high-voltage silicon carbide platform, including ultra-fast charging technology and energy storage, which will significantly increase costs. Nio and Tesla are aggressively deploying this, without considering matching it with current sales. How does Xpeng judge return on investment?

(No response provided in the original text)# Translation

何小鹏: I will consider the rhythm and cost, as well as the cost of technological changes and replacements. After doing it for a while, there will be a lot of changes when the technology is replaced.

Moreover, I will consider the changes in the stages of intelligent assisted driving in 2023, 2025, and 2027, and consider many rhythms. But I have set a goal that we must do well in energy efficiency by 2023. This is something we firmly believe in, not just in China, but also overseas.

On Market Share Expectations

Question: Today, XPeng stated that you believe the penetration rate in 2025 will be close to 50%, whereas previously you had estimated 35%. This is a significant change. What do you think XPeng’s market share will be by 2025?

何小鹏: If we assume that a hypothesis is established, for example, from 2024-2026, the Chinese automotive field will become like the Warring States Period, with many fighting against each other, and only a few remaining. At that time, we will know who the front-runners are and who is struggling to survive. In order to enter the Warring States, a significant penetration rate must be achieved. Some of the ambitious penetration rate targets proposed by some companies are not ambitious from that time’s perspective.

Question: Do you think XPeng should also have such a market share?

何小鹏: To enter the Warring States, the front-runners must also have such a market share. Today’s situation is too dispersed.

Question: Xiaomi has also entered the market, and traditional automakers are also transitioning, including BBA, which has very aggressive targets. If the head companies can achieve such market share.

何小鹏: Everyone has a chance and a challenge, but for us, we are much smaller than they are, so we are forced to think farther, under greater pressure, and run faster.

On the Allocation of Personal Energy by He XPeng

Question: The competition in the automotive main track itself is also fierce, so with so many ecological corporate frameworks, how do you allocate your personal energy?

何小鹏: I allocate my energy dynamically. Recently, I have spent more time in the field of robotics. But the vast majority of my time is still in the XPeng automotive system because it is a large strategic layout. What everyone sees today is difficult to associate with XPeng, but in three to five years, XPeng’s work will have more links with what they will be doing four to five years later, and then you will see things that are 10 times cooler than today.

On Hardware Upgrades

Question: Will our future car models consider hardware compatibility? Some competitors provide upgrade capabilities on the basic chips. Will we have this?

何小鹏: We, like our competitors, have basic upgrade capabilities, but these are not of much significance for higher-level intelligent assisted driving like XPILOT. Looking at a complete vehicle, such as control redundancy and steering redundancy, these cannot be upgraded. You can only upgrade some computing power, which is valuable for the car but not comprehensive.### Thoughts on self-developed and contract manufacturing models

Question: XPeng has highlighted the concept of being full-stack, but the industry still believes that professional tasks should be left to professionals. Many companies are researching smart technologies and develop white-label products for automakers. In XPeng’s opinion, which model is the primary one?

He XPeng: There’s more than one path to success.

In times of rapid and large-scale innovation, pursuing efficiency and change requires self-development. However, the more stable the industry becomes, the clearer the division of labor and stronger globalization. Therefore, it is best not to choose self-development, as you cannot match the scale, technology, or management of others. Instead, you should turn to them. If we develop something ourselves but another company can do it better or more cheaply once it becomes stable, we should let them do it, freeing up our time to focus on more valuable tasks.

From my perspective, the changes the auto industry will face by 2024, and even by 2026, will exceed our imagination.

Even entrepreneurs like us at the forefront cannot predict how quickly these changes will occur. If partners cannot keep up because they need to grow and develop, they will inevitably be slower than us. This is due to market logic. They might have stronger capabilities, but they don’t need to think about it from our perspective. However, if we stabilize, we may open up some opportunities.

About autonomous driving

Question: We currently see automatic parking, NGP for high-speed and city driving. In your opinion, when will these three motor skills become fully established?

He XPeng: It depends on the quality of implementation. I think there will be better implementation by next year. A well-established implementation will be around 2025-2027. A complete implementation may not happen until after 2030.

Reporter: WM Motor has pioneered the first autonomous driving manufacturer to assume responsibility when accidents happen. What is your opinion on this?

He XPeng: Only the legislator can answer this legal question, not a civil one. The most important thing is to ask those in charge of the regulatory authorities to speak, so we will have to wait and see.

Question: Is Robotaxi operating P5 or customized models?

He XPeng: We will have another vehicle next year, which may be P5 or another model. We are still open-minded.

Question: In the second half of 2022, XPeng would like to become the first Chinese automaker to explore autonomous driving operations. Does this refer to Robotaxi or other plans? I remember that G3 used to operate Aixing Travel. What does that mean?# Markdown text in English with HTML tags

何小鹏（XPeng He）: The core logic of our system is similar to what you mentioned earlier – we operate within a specific range of a city with safety personnel on board. The primary value of our system lies in the accumulation of data― especially the exceptional data that we can gather, such as road closures or handling heavy rain conditions. It is only through accumulating and analyzing this data over time, by driving millions of kilometers, that we can lay the groundwork for future autonomous driving and unmanned driving in larger areas.

Ultimately, the key factor is robustness. Increasingly, we are finding that our unmanned driving technology needs to have the ability to operate smoothly in a variety of environments and scenarios. While it is relatively easy to run simulations along a set path, a truly robust system must be able to handle any situation. New Energy’s XPILOT 5.0 tackles this challenge with two fully integrated loops that work together to process large amounts of data. Without this kind of continuous data accumulation, achieving full L4 or L5 automation is very unlikely.

Question: How do we achieve the safety redundancy and user experience of Robotaxis? What sets us apart from L4?

何小鹏: I think some Robotaxis lack safety redundancy, or they do not have enough redundancy built in – this can either be in software or hardware. Some Robotaxis have relatively good safety redundancy, but many others do not, especially when it comes to expensive hardware. With XPILOT 5.0, we have put significant effort into achieving redundancy and safety from a hardware perspective. Companies that achieve L4 level functionality often still need to invest in hardware and algorithms to succeed. In other words, hardware, algorithms, and data are needed to achieve redundancy and safety, and companies still need to make much progress in this direction, including us at Xpeng.

Thinking about Future Development

Question: Today, we are seeing Xpeng expanding into many different product categories. Is there any plan for Xpeng to delve more deeply into fundamental technologies within the auto industry? When we talk about “full stack”, we often focus only on software. Do you have any ideas about hardware or lower-level technology?

何小鹏: Today, talking about full stack hardware development in the auto industry is daring. I must say that hardware refers to everything, including radar calculation and camera sensing. We have done a lot of hardware development in the design of printed circuit boards (PCBs) and domain controllers, and this is just the beginning. In the next two years, you’ll see us do a lot of interesting work, although what exactly that will comprise is yet to be seen.

Question: Currently, people often say that NIO has stronger service capabilities, while Xpeng has stronger AI technology. However, I believe that these differences are merely tactical in nature, and in the future, both companies will gravitate toward one another in terms of their respective strengths. Their core competitive advantages will still differ, however.

何小鹏: I agree, and I think that service and technology are very much interconnected. Achieving a human-like experience and being able to provide excellent service is highly dependent on having strong AI capabilities. In the future, I think that all companies will have to strive to develop a meaningful, unique advantage in this fast-evolving industry.# Xiao Peng: My understanding is slightly different. I think the essence of a company is composed of both its form, which is what you see on the surface, and its core, which is the essence embedded in its bones. Xiao Peng’s core is a technology company, and what he sees outside is only a small part of what is actually there. The cultural values and thinking inherited deep inside a company’s bones, which determine its future Chinese and global strategies, are the company’s DNA. Some methods may reflect in quicker and better results, while other DNA may result in slower but more comfortable long-term benefits. These are different cultures, different paths.

Question: Some car companies construct grand scenes, while we have seen Xiao Peng release many technologies but not implemented them into a fully integrated product. How can we reflect Xiao Peng’s tone on intelligent technology branding? Can we show consistency?

Xiao Peng: We take a different approach. Our logic is progressive. You can think far and fancy, but the rhythm is the most important. There are two options for rhythm. One is to jump directly from the initial stage to the ultimate, and the other is to proceed in five or six steps, in which we favor the latter more.

In my opinion, the majority of macro-level thinking regarding innovation is correct, but the micro-level thinking is often wrong. You want to see the faraway goal, like longitude and latitude, but you do not know the exact location. You need to continually adjust your direction in the process of moving forward. It may only be a 3% or 5% adjustment, but it is essential. That is the process of entrepreneurship.

Question: For Xiao Peng himself, what external and internal factors influence the vertical integration, high-level integration, and professional division of labor? When might he outsource in core fields, and when might he collaborate on vertical and professional division?

Xiao Peng: The core, the secondary, and the unimportant parts of a company are like different rings inside a hurricane. The closer to the center of the hurricane, the more investment and time we put in, but some abilities may have moved away from the center after five years because they stabilized without much change.

Therefore, we have a technical analysis meeting every six months to discuss whether a technology will continue to have value in two or three years. If we cannot see any value, we reduce investment, or at least maintain a stable level of investment. If we see many possibilities, we increase investment. If we stop investing, we might find that two or three other things it sparks off also need investment, and then we will strengthen our investment. It is a dynamic process.

If someone else can do it better and cheaper than us, and we cannot see that it is a core point in our strategy, we should cooperate with them from a long-term perspective. This is more likely to lead to success. Giving up something, Xiao Peng gains something, and we also have to give something up, or the business will collapse.

Decision-Making Thinking about Exploratory IndustriesQuestion: XPeng Motors released Machine Horses and flying cars this year which look pretty cool. However, these products face commercialization and landing issues in their respective fields. Is XPeng Motors really prepared to make money from this? How do they consider the input and output?

He XPeng: First, they are all XPeng’s ecological companies. Any company must have enough profits in the long run. He needs to give back to shareholders and employees in order to bear his R&D costs.

We consider strategy as the main focus when we first started doing ecological companies, while financial is the main focus of investment companies, so we only consider strategy first now.

Question: XPeng is a company that likes to explore. It has done roof cameras in XPeng G3, but such designs have not been seen in subsequent models, including today’s flying cars and machine horses. From your personal point of view, why does XPeng want to try innovative products and functions? Do you have any thoughts on the impact of failed exploration on the company?

He XPeng: If there is exploration, there will definitely be failures. If there is no failure, it is not exploration. This way, it can be industrialized or commercialized.

Secondly, I think for more and more entrepreneurs in China, they want to do better things rather than make more money. They put this in their company’s products and their corporate culture.

When I first invested in XPeng Motors, I sat in a car in 1989 for the first time. The car back then was not far from the car now, and I haven’t driven a new one.

Why hasn’t there been a huge change in cars in 30 years? If we go make another car that doesn’t have a huge difference, just pursues its design, positioning and price differences, what am I doing in entrepreneurship? Entrepreneurship is about doing different things, like every one of us being a different person.

Question: You mentioned that the flying car price will go up to the range of 1 million yuan. Do you need to establish a sub-brand to take over?

He XPeng: XPeng Huitian is not the same brand as XPeng.

Question: That car will be Huitian’s product, not XPeng’s?

He XPeng: It will be sold under XPeng’s ecological system, and you can think of XPeng as its dealer.

Thoughts on ecology

Question: Now Tencent is also saying that it will not participate in car making, but only build a network. Huawei is also saying the same thing. The other day, I heard that Xiaomi’s biggest advantage is its ecological advantage, and all the household appliances are replaced by Xiaomi. Today, listening to you, XPeng will also build its own complete ecology. When everyone has different ecosystems, where is the collision point of competition? How to highlight XPeng’s own advantages?

He XPeng: First, I think XPeng is still a relatively small company, far from the platform they said, one net, or three-dimensional ecology that I really see. But they are all creating a strong network effect, not an ecological effect.- Second, we expect to have our own ecosystem in the future, but it will take 5-15 years, and maybe we can share it later.

• However, our most important task today is to deepen and improve the capabilities of AI and manufacturing, gradually make China and the world better, this is our core task. Each company has its own strengths and weaknesses, but it doesn’t matter. The world’s technological changes really happen very quickly, and a few years ago we thought that no one could break through certain companies in the social field, maybe in a few years, the probability is very low, but there is definitely a possibility, this is the charm of technology.
• Q: Why can you accept such a high valuation for XPeng Motors? From Pengxing to Zhineng Huitian and other ecosystem companies, can you talk about XPeng’s thinking on ecology?
• He XPeng: Huitian or Pengxing are our ecosystem companies. We may have light investments in these companies in the future. We hope to do it based on XPeng’s strategic and product positioning. Secondly, XPeng will help them with manufacturing and part of the sales, so their sales scale and manufacturing requirements will be improved. In these companies, XPeng Motors and myself will invest, so that they have more sufficient funds. Huitian recently announced more than 500 million US dollars in Series A financing, which was mainly invested by China’s top VC and PE.
• Q: Will you invest in other fields in the future?
• He XPeng: We will share it with you when we have done it and achieved certain results.
• Q: Xiaomi has done a great job in the hardware ecosystem, can you talk about your thinking and similarities or differences with Xiaomi?
• He XPeng: I think Xiaomi has done a great job in the hardware ecosystem, but it is difficult in the field of transportation. Therefore, we will take a new path that is different from theirs. We will share it with you when we open up the situation in a few years.
• Q: We have demonstrated the field of robotics, flying cars, and you mentioned industrial manufacturing before. How has XPeng done in software, including large-scale algorithms and various capabilities? When we talk about machine horse autonomous driving, are we still self-developing their algorithms? How has XPeng Motors done in this area?
• He XPeng: Our robot’s autonomous driving is separated from XPeng, but it has XPeng’s code, otherwise it cannot be done so quickly, and it must stand on the shoulders of giants to do so.
• Q: Will there be interactions on key hardware in the future?
• He XPeng: There are many, such as power and materials science.
• Easter Egg:
• Q: What do you think are the characteristics of programmers?Question: Today’s event was quite wonderful. The first question is presented in the form of a debate. I would like to ask you this question again: If you could be sent to a society with more advanced technology 50 years from now, would you like to go? Why?

He XPeng: In my current state, I do not want to go. I want to live to 50 years from now and do not want to be sent there.

This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.